JOURNAL OF APPLIED ECONOMETRICS, VOL. 5, 29-46 (1990) INTERPRETING AN ERROR CORRECTION MODEL: BEHAVIOUR, AND DYNAMIC INTERNATIONAL MONEY DEMAND zyxw PARTIAL ADJUSTMENT, FORWARD-LOOKING IAN DOMOWITZ zyxwv Department of Economics, North western University, 2003 Sheridan Road, Evanston, zyxw IL 60208, USA AND CRAIG S. HAKKIO Economic Research Department, Federal Reserve Bank zyxwv of Kansas Cily, Kansas City, zyxw MO 64198, USA SUMMARY An error correction model is derived from a stochastic dynamic programming problem incorporating rational expectations. A parametric restriction is derived that allows a test for the theoretical proposition that the optimal strategy behind the error correction form entails the failure to asymptotically close the gap between the choice variable and the growing target. This is accomplished by nesting a partial adjustment model with forward-looking expectations within the error correction paradigm. The counterintuitive behaviour embodied in the error correction model is not supported by the data in the context of a cross-country comparison of cash balances relationships. 1. INTRODUCTION The error correction model has become a very popular specification for dynamic equations in applied economics, including applications to such mainstream problems as personal consump- tion, investment, and the demand for money. zyxwv ’ The statistical framework is attractive, in that it encompasses models in both levels and differences of variables and is compatible with long- run equilibrium behaviour. The success of the error correction paradigm in applications has led to the development of theory justifying the form of such an estimating equation for purposes of interference (i.e. the concept of cointegration in economic time-series-Granger and Engle, 1988, and related literature), as well as discussion of the theoretical behaviour of such models under so-called ‘growth equilibrium’. We contribute to the latter discussion in this paper in an empirical context, through a cross- country study of the demand for real money balances. The major deficiency of the error correction approach in practice has been the lack of an underlying optimization model that generates testable restrictions within the error correction framework and allows ‘structural’ ’ Discussion and references are contained in Hendry, Pagan, and Sargan (1984). More recent work on money demand, in particular employing the basic concept, includes that of Rose (1985); Baba, Hendry, and Starr (1987); Dornowitz and Elbadawi (1987); and Klovland (1984). ‘See, for example, Currie (1981). Salmon (1982), and Nickel1 (1985). 0883-7252/90/010029- 18$09.00 zyxwvu 0 1990 by John Wiley & Sons, Ltd. Received February 1987 Revised August 1989